Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR Ivan - PowerPoint PPT Presentation

Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR Ivan Bazarov, William Bergan, Cameron Duncan, Acknowledgements: David Rubin, Jim Sethna DOE DE-SC0013571 Cornell University

CESR

Sloppy Models Problem Statement  Minimize one objective (vertical emittance/beam size)  Large number of decision variables (independent magnets)  No reliable auxiliary information (dispersion, coupling)  Reliable model of machine responses (BMAD simulation)

Sloppy Models Phys. Rev. E 68 (2003) 021904.

Simulated Results

Experimental Results RCDS - Nucl. Instr. Meth. 726 (2013) 77.

Conclusions  Knobs provide some improvements  Still far from quantum limit  Something missing from models?

Multi-Objective Genetic Algorithm Problem Statement  What if: ● Competing criteria of optimal machine performance ● In regime where model of machine responses is unreliable  Needed: a model-agnostic search for optimal performance trade-offs

c 2 , 1 0, 1 0 xy x y 1 y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 x dominates o y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 neither x nor o dominates y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 y c 0 0 c 1 x

c 2 , 1 0, 1 0 xy x y 1 y c set of non-dominated points 0 0 c 1 x

genetic algorithm (spea2) toy example 1 parent population Objective B How it works 0 0 1 Objective A

genetic algorithm (spea2) toy example 1 offspring Objective B 0 0 1 Objective A

genetic algorithm (spea2) toy example 1 Objective B survivors (parents of the next generation) 0 0 1 Objective A

● Needed: a model-agnostic search for optimal performance trade-offs ● We tested an elitist genetic algorithm with re- sampling on bdad simulations of CESR ● Solution set shows randomness but converges in statistics ● Numerical evidence that power-law fit to solution set is an unbiased estimate of trade- off front

Preliminary Results

Final Thoughts  Any real-life online optimization metaheuristic is likely to be a combination of model-cognizant and model- agnostic parts;  Machine safety needs to “filter” trial solutions preventing them from adopting forbidden states;  Noise handling and maximizing throughput are always key issues;  CESR is an ideal platform to deploy new kinds of online optimization strategies, including AI and stochastic algorithms.